1. BMS 524 - “Introduction to Confocal Microscopy and Image Analysis”
Lecture 4: Image file formats
Department of Basic Medical Sciences,
School of Veterinary Medicine
Weldon School of Biomedical Engineering
Purdue University
J. Paul Robinson, Ph.D.
SVM Professor of Cytomics
Professor of Biomedical Engineering
Director, Purdue University Cytometry Laboratories,
These slides are intended for use in a lecture series. Copies of the slides are distributed and students encouraged to take their notes on these graphics. All material copyright J.Paul Robinson unless otherwise stated. No reproduction of this material is permitted without the written permission of J. Paul Robinson. Except that our materials may be used in not-for-profit educational institutions ith appropriate acknowledgement. It is illegal to upload this lecture to Course Hero or any other site.
You may download this PowerPoint lecture at
This lecture was last updated in January, 2018
Find other PUCL Educational Materials at

2. What are data file formats?
Data files are a representation of an original image, which is itself a representation of reality.
The chain of digital image processing includes both creation of digital data from an image, and recreation of an image from the digital data.
Data file formats are created in order to make specific operations more convenient. The most convenient format may differ with the particular application.

3. Image process - from subject to output
digital
Analogue

4. How it works in real life

5. What is image manipulation?
For most purposes, a one-to-one mapping of pixels to data values is most useful, but the internal representation of the data values may be different for different file formats.
Files can be either compressed, or not, and compression can be either lossy or not. For scientific analysis lossy compression is unacceptable; it may be useful for overview presentations.
Image manipulation can take place before image acquisition, during image acquisition, on the digital data, or during recreation of an output image.

6. 8 bits provide
256 possible
gray levels

7. How does this translate from an image into an image map?

8. How does this translate from an image into an image map?
512
512
512
512
An image need only be parsimonius, i.e., it only needs to show what is necessary to provide the expected image.

9. It depends on what you are collecting the image for as to the resolution required

10. How does this translate from an image into an image map?
min
0=black, 255=white
max

11. Pixel values are in a list file

12. Pixels
512
255
512

13. Image size
The higher the resolution of the image, the more data points there are
Very high resolution files need to be reduced in size to store the data
We need to employ compression algorithms to reduce the file size
But the goal is to maintain the quality of the image
Lossless compression is ideal for biological images

14. Compression
Run Length Encoding – count the number of identical values, replace the values with a count followed by the value
A kind of compression algorithm which replaces sequences ("runs") of consecutive repeated characters

15. How big is an image? Size is dependent on several factors –
The X-y size in pixels
The number of bits (colors)
For example, a 640 x 480 x 24 bit image = 7,372,800 = 921,600 bytes
The same image “double the size” (4 times pixels)
1280 x 960 = x 24 bits = 29,491,200 = 3,686,400 bytes
An image at full MacThunderbolt screen at
12560 x 1440 = 3,686,400 x 24 = 88,473,600 bytes

16. 0% Compression
50% Compression
76k
42k
75% Compression
80% Compression
28k
26k

17. The 76 k file
The 26 k file

18. 4 x magnification
The 76 k file
The 26 k file

19. What is image analysis?
Brightness and contrast variation are controlled by a system input-output curve.
Spatial kernel filtering and median filtering use information local to a particular area of an image to modify that area.
Digital image analysis is “Data Analysis”.

20. Original Image

21. Contrast enhanced image

22. Both images

23. Conversion of image manipulation into new image

24. Smoothing

25. Smoothing Filter
1 1 1

26. Smoothing
9 x 9 Smooth
3 x 3 Smooth
original

27. Sharpening
9 x 9 Sharpen
3 x 3 Sharpen
original

28. Gaussian Filter
1 2 1
2 4 2
Gaussian filters are filters where a filter is a filter whose impulse response is Gaussian – it is designed to give no overshoot to a step function input while minimizing the rise and fall time

29. Sobel Filter
1 2 1
0 0 0
Sobel is a discrete differentiation operator, computing an approximation of the gradient of the image intensity function. It basically operates in the vertical and horizontal planes and thus produces edges (abrupt changes in gradient)

30. Application of a median filter
A median filter is often desirable to be able to perform some kind of noise reduction and it preserves edges while removing the noise

35. Comparing Images (Thumbsplus) file size: 73,011 file size: 21,358
Copyright J. Paul Robinson, Steve Kelley Purdue University Cytometry Laboratories ‹#› of 42 Slides
file size: 73,011
file size: 21,358

36. It depends on what the purpose is of your image as to the resolution needed
file size: 73,011
file size: 21,358
File size 12,470
An image need only be parsimonius, i.e., it only needs to show what is necessary to provide the expected image.

37. What are you loosing?
Same image as previous page, expanded times
Copyright J. Paul Robinson, Steve Kelley Purdue University Cytometry Laboratories ‹#› of 42 Slides

38. Some image formats (Lossy)
Approximately 1382 different image formats
RAW – this differs from product to product so there is no standard
PNG (Portable Network Graphics) uses ZIP compression which is lossless, and slightly more effective than LZW (slightly smaller files) (designed for web).
More recent format, royalty free, note patent issues over the LZW format (GIF).
GIF files can be saved with a maximum of 256 colors. (not good for photographic images)
JPEG (JPG)— Joint Photographic Experts Group
PSD, PDD - Photoshop's native format. PSD preserves layers, transparency, adjustment layers, masks, clipping paths, layer styles, blending modes, vector text & shapes
BMP is an uncompressed proprietary format invented by Microsoft.
Copyright J. Paul Robinson, Steve Kelley Purdue University Cytometry Laboratories ‹#› of 42 Slides

39. Lossless Formats ILBM – (lossless RLE compression of Amiga IFF images)
JBIG2 – (lossless or lossy compression of B&W images)
JPEG-LS – (lossless/near-lossless compression standard)
JPEG 2000 – (includes lossless compression method)
JPEG XR – formerly WMPhoto and HD Photo, includes a lossless compression method
PGF – Progressive Graphics File (lossless or lossy compression)
TIFF – Tagged Image File Format (LZW compression, lossless data-compression)
Gifsicle (GPL) – Optimize gif files
Jpegoptim (GPL) – Optimize jpeg files
Copyright J. Paul Robinson, Steve Kelley Purdue University Cytometry Laboratories ‹#› of 42 Slides

40. Lossy Compression Most filters will result in lossy compression
This means once you compress the image, you can never return the image back to its original self
One needs to be careful with biological data to ensure that you preserve the raw files
Lossless: GIF, TIF, JPEG 2000
(GIF is usually limited to an 8-bit palette, or 256 colors and a single transparency color)
Lossy: jpg
Copyright J. Paul Robinson, Steve Kelley Purdue University Cytometry Laboratories ‹#› of 42 Slides

41. Vector Image file formats
EPS (Adobe Encapsulated Postscript)
AI (Adobe Illustrator)
PDF (Portable Document format
SVG – W3C standard vector image format (Scalable Vector Graphics) (Inscape and Corel Draw)
DXF – Drawing eXchange Format (CAD format from Autodesk)

42. A vector image (Corel Xara Design)
In a vectored image, every component is editable – whereas in most other images, unless it is composed of multiple layers, then editing that image is quite difficult.
Most cell images formats will not be vector images.

43. Summary Structure of data files Image manipulation Kernels Filters
Lossy and Lossless compression